How the tool works¶
Each screen has one job
The experience separates processing, browsing, detailed review, and portfolio learning so the steps do not stack into one endless page.
1. Process a claim¶
Choose an existing claim or upload a claim document. The tool reads the file and creates a new processing run. Previous runs stay available for audit history, but the normal workflow shows the latest result.
Example: A 40-page flood claim is processed. The tool flags possible tree debris and special-limit issues and records the pages where each appeared.
2. Browse the Claims Library¶
The library shows one row per claim. A reviewer can search by claim or policy, filter by review status, and narrow the list to a specific violation type.
3. Review the evidence¶
The Claim Review workspace keeps the original document, the processing summary, and a fixed-height issue table together. For each issue, the reviewer can:
inspect its page and description;
rate it from one to five stars;
confirm or reject it; and
leave a concise explanation for the next reviewer.
Edits save to the shared data store, allowing reviewers to collaborate without copying results into separate spreadsheets.
4. Learn across claims¶
Violation Analytics counts how many distinct claims contain each violation type. This avoids making a repeatedly processed claim look like many separate claims.
Selecting a type reveals the affected claims, pages, ratings, and review outcomes. High rejection rates can point to unclear rules or overly sensitive detection logic.